Nanomechanical detection of antibiotic mucopeptide binding in a model for superbug drug resistance JOSEPH WAFULA NDIEYIRA1,2#, MOYU WATARI1#, ALEJANDRA DONOSO BARRERA1, DEJIAN ZHOU3,4, MANUEL VÖGTLI1, MATTHEW BATCHELOR3, MATTHEW A. COOPER5, TORSTEN STRUNZ1, MIKE A. HORTON1, CHRIS ABELL3, TREVOR RAYMENT6, GABRIEL AEPPLI1 AND RACHEL A. McKENDRY1*. 1. London Centre for Nanotechnology and Departments of Medicine and Physics, University College London, 17-19 Gordon Street, London WC1H 0AH, UK 2. Jomo Kenyatta University of Agriculture and Technology, Department of Chemistry, PO Box 62000, Nairobi, Kenya 3. Department of Chemistry, Lensfield Road, University of Cambridge, Cambridge CB2 1EW, UK 4. School of Chemistry and Astbury Centre for Structural Molecular Biology, University of Leeds, Leeds LS2 9JT, UK 5. Institute for Molecular Bioscience, The University of Queensland, Brisbane, 4072, Australia. 6. School of Chemistry, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK # These authors contributed equally * email: [email protected]
(Dated: 12 October 2008)
The alarming growth of the antibiotic-resistant superbugs methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant Enterococcus (VRE) is driving the development of new technologies to investigate antibiotics and their modes of action. We report the label-free detection of vancomycin binding to bacterial cell wall precursor analogues (mucopeptides) on cantilever arrays, with 10 nM sensitivity and at clinically relevant concentrations in blood serum. Differential measurements quantified binding constants for vancomycin-sensitive and vancomycin- resistant mucopeptide analogues. Moreover, by systematically modifying the mucopeptide density we gain new insights into the origin of surface stress. We propose that stress is a product of a local chemical binding factor and a geometrical factor describing the mechanical connectivity of regions affected by local binding in terms of a percolation process. Our findings place BioMEMS devices in a new class of percolative systems. The percolation concept will underpin the design of devices and coatings to significantly lower the drug detection limit and may also impact on our understanding of antibiotic drug action in bacteria. When biochemically specific interactions occur between a ligand that detects mass related changes in the dielectric constant10 13. immobilized on one side of a cantilever and a receptor in Cantilevers are therefore unique as probes of small molecule solution, the cantilever bends due to a change in surface stress1 9. drug binding interactions and, by virtue of their miniaturized The general applicability of this novel nanomechanical dimensions they are amenable for parallelization14,15 for high biosensing transduction mechanism has been shown for throughput screening of thousands of drugs per hour. sequence specific DNA hybridization1 5,8, single base Here we report the first quantitative differential mismatches1, DNA quadruplex5, protein recognition1,3,7,9 and nanomechanical investigation of drug target binding interactions recently the detection of interferon alpha induced I 8U gene on multiple cantilever arrays focusing on the antibiotic expression in total human RNA, a potential marker for vancomycin (Figure 1). Today vancomycin is one of the last melanoma progression and viral infections8. However, to date, powerful antibiotics in the battle against resistant bacteria and multiple cantilever arrays have not been applied to quantify the ‘hospital superbug’ MRSA16 27. It is a vital therapeutic drug drug target binding interactions, despite offering considerable used worldwide for the treatment of infections with Gram advantages. First, cantilevers require no reporter ‘tags’ or positive bacteria, particularly those Staphylococci and external probes and so biomolecules can be detected rapidly in a Enterococci responsible for post surgical infections. single step reaction. Second, cantilever arrays can screen Vancomycin binds to the C terminus of peptidoglycan multiple drug target interactions and reference coatings in mucopeptide precursors terminating in the sequence Lysine D parallel and under identical experimental conditions. Third, we Alanine D Alanine16 18,20,21 as shown in Figure 1. This have previously shown that quantitative ligand receptor binding interaction blocks the action of bacterial transpeptidases and constants can be measured on cantilever arrays2. Moreover, the transglycosylases, which catalyse the cross linking of the nanomechanical signal is not limited by mass, in contrast to growing bacterial cell wall, resulting in cell lysis16 27. evanescent techniques such as surface plasmon resonance (SPR) Unfortunately, due to the over use of antibiotics,
1
Figure 1 The nanomechanical detection of vancomycin-mucopeptide analogue interactions on multiple cantilever arrays. (a) Schematic diagram to show cantilevers coated with DAla (vancomycin sensitive), DLac (vancomycin resistant) or PEG (reference) alkanethiol SAMs. Vancomycin is injected in solution and binds specifically to the mucopeptide analogues causing the cantilever to bend downwards due to a compressive surface stress. (b) The chemical binding interaction between vancomycin and the bacterial mucopeptide analogue, DAla. It is known from solution phase studies that the specificity of this complex arises due to (i) the interaction of the C terminal free carboxylate of the peptide with three amide bonds in the vancomycin backbone (ii) the formation of two C=O H N hydrogen bonds and (iii) hydrophobic interactions of the alanine methyl groups with aromatic residues of vancomycin. The dashed lines represent the 5 intermolecular hydrogen bonds. The yellow dashed line represents the hydrogen bond associated with bacterial resistance; (c) The deletion of a single H bond in mutated DLac mucopeptides gives rise to drug resistance. The binding pocket of vancomycin is represented schematically and the grey dotted line represents the deleted hydrogen bond and electrostatic repulsion between the oxygen lone pairs of electrons. resistance to vancomycin is rapidly increasing and now poses and functionalized with alkanethiol self assembled a major international public health problem22, 24,26. Bacterial monolayers (SAMs) of (i) the drug sensitive mucopeptide resistance in Enterococci can arise due to the subtle change of analogue HS(CH2)11(OCH2CH2)3O(CH2)(CO)NH(CH2)5(CO) an amide linkage to an ester linkage in the growing bacterial L Lys (ε Ac) D Ala D Ala, herein termed DAla; (ii) the cell wall, resulting in the deletion of a single hydrogen bond mutated sequence which confers vancomycin resistance in from the binding pocket, rendering the antibiotic VanA and VanB resistant Enterococcal phenotypes, 16 27 therapeutically ineffective (Figure 1). The development HS(CH2)11(OCH2CH2)3O(CH2)(CO)NH(CH2)5(CO) L Lys of novel methods to detect and quantify the binding of (ε Ac) D Ala D Lac, termed DLac. Our previous studies2,5,6,8 antibiotic–mucopeptide interactions is thus of significant have emphasized the importance of acquiring differential clinical importance. In addition the structure and binding measurements using a reference cantilever and here we use a properties of vancomycin mucopeptide complexes have been cantilever coated with an ‘inert’ SAM terminating in 16 27 extensively studied both at surfaces and in free solution triethylene glycol HS(CH2)11(OCH2CH2)3OH termed PEG, and thus serve as a model system to evaluate the capabilities which is known to resist biomolecule adsorption on of cantilevers in small molecule drug target detection. surfaces28 30. The Supplemental Material describes the synthesis of DAla, DLac and PEG. The absolute deflection at DETECTION OF VANCOMYCIN MUCOPEPTIDE INTERACTIONS the free end of each cantilever zabs was measured using a time multiplexed optical detection system in different buffer To probe the in plane nanomechanics of antibiotic drug and blood serum environments under constant flow. The target interactions, multiple arrays of eight rectangular silicon bending signal was subsequently converted into a differential cantilevers were coated on one side with a thin film of gold surface stress between the upper and lower sides of the
2 Figure 2 Investigating the specificty and sensitivity of antibiotic-mucopeptide interactions on cantilever arrays. (a) Absolute bending signal of DAla1 (red), DAla2 (orange), DLac1 (light blue), DLac2 (dark blue), DLac3 (dark green) and in situ reference cantilevers PEG1 (black) and PEG2 (grey) coated cantilevers to phosphate buffer, 250 M vancomycin, and return to phosphate buffer. A negative signal corresponds to a compressive surface stress and the downwards bending of the cantilever, as illustrated in Figure 1a; (b) The corresponding differential bending signals of DAla1 (DAla1 minus PEG1, shown in red) and DLac1 (DLac1 minus PEG1, shown in blue); (c) Differential DAla signals for 10, 100, 1000 nM vancomycin. The differential PEG reference signal is shown (PEG2 – PEG1 black); (d) Differential signals of three DAla cantilevers for 10 nM vancomycin. The differential PEG reference signal is shown (black).
31 cantilever σabs, using Stoney’s equation The nanomechanical force exerted by vancomycin 2 peptide interactions was investigated on microfabricated 1 t E cantilevers. The deflection of an array of cantilevers coated σ abs = zabs (1) 3 L 1−ν with DAla, DLac or in situ reference PEG SAMs, was monitored in parallel upon injection of different where L is the effective length of the cantilever up to 500 m, t concentrations of vancomycin in sodium phosphate buffer is the thickness ~ 0.9 m, and E 1( −ν ) = 180 GPa is the (pH 7.4, 0.1 M). Typical absolute bending signals for one ratio between the Young’s modulus E and Poisson ratio ν of array comprising two DAla, three DLac and two PEG coated Si (100)32. We used a home built gravity flow system to control cantilevers are shown in Figure 2a. In buffer, we observed the exchange of up to six different vancomycin solutions (10 nM that all cantilevers showed a stable baseline. Upon injection – 1 mM), 100 mM sodium phosphate buffer and 10 mM HCl of 250 M vancomycin, both DAla1 and DAla2 coated regeneration solutions, via an automated valve (Supplemental cantilever rapidly bent downwards (illustrated in Figure 1a), Material). reaching a stable equilibrium absolute compressive bending The aim of our investigations was to ascertain whether signal of 180 nm and 172 nm respectively after 30 mins cantilever arrays have the sensitivity to quantify vancomycin– under constant flow conditions. Conversely the DLac DAla binding interactions and detect the deletion of a single cantilevers showed a small absolute downwards bending hydrogen bond associated with antibiotic resistance to the mutated peptide analogue, DLac. Moreover, we examined the signal of 38, 31 and 31 nm for DLac1, DLac2 and DLac3, sensitivity of antibiotic detection in blood serum at clinically respectively. The two reference PEG coated cantilevers, relevant concentrations33 of 5 40 g/ml, which corresponds to 3 PEG1 and PEG2 showed a small downwards bending signal 27 M. In addition, we sought to alter the surface peptide of 14 nm and 13 nm. Upon injection of the buffer, the density in order to optimize drug detection sensitivity and to signals were observed to converge back towards the stable investigate the underlying mechanotransduction mechanism. ‘zero stress’ baseline. This step was then followed by 10 mM Nanomechanical biosensors can best be exploited only if we HCl, which is known to dissociate vancomycin peptide develop a fundamental understanding of what causes the complexes and regenerate the free peptides for further 13 cantilever to bend and this knowledge will aid the development antibiotic studies. of new devices with significantly improved drug detection It is known that the absolute bending signals are a sensitivity. convolution of biologically specific binding events and non-
3
Figure 3 Nanomechanical detection of antibiotics in blood serum at clinically relevant concentrations. Differential bending signal of DAla1 (red) and DLac1 (blue) upon injection of 7 M vancomycin in 90 % fetal calf serum and 10% sodium phosphate buffer at pH 7.4. specific influences, including reactions occurring on the underside of the cantilever, liquid injection spikes, changes in refractive index and temperature1,2,5 8. These non specific effects will affect DAla, DLac and PEG signals to the same extent, and are therefore removed by taking a differential measurement using a reference cantilever with an inert coating.2,5 8 The differential measurements, shown in Figure 2b revealed the surface forces induced by biochemically specific vancomycin peptide interactions. Upon injection of 250 M vancomycin, the differential surface stress signal for DAla1 (DAla1 PEG1) and DLac1 (DLac1 PEG1) were Figure 4 Nanomechanical drug-target percolation on cantilever found to be 35.3 mN/m and 5.1 mN/m respectively. arrays. (a) A three dimensional graph showing the measured To examine the reproducibility of the nanomechanical differential surface stress response for DAla (red circles) and DLac signals, we performed more than 100 measurements on four (black circles) coated cantilevers as a function of vancomycin different cantilever arrays, each composed of eight concentration in solution [Van] and DAla surface coverage p, cantilevers. The raw bending signals were analyzed using superimposed with the results of the global fit according to Equation 2; (b) Least squares analysis to determine the best values for p and automated data software developed in our group to rapidly c α.. Plot shows the chi squared values (blue) and α (red) for each examine large data sets and remove bias6. The average value of pc. The black, blue, and red arrows indicate the chi squared nanomechanical surface stress signal for 250 M vancomycin minimum, pc, and α values at the chi squared minimum, on one array was 34.6 ± 0.9 mN/m for DAla and 4.2 ± 0.5 respectively. mN/m for DLac. The mean signals across four arrays was 34.2 ± 5.9 mN/m for DAla and 3.8 ± 1.5 mN/m for DLac. The high reproducibility of within-array measurements, and an increased variance associated with between-array measurements, agrees with our previous findings6. defined as the DAla surface coverage fraction on the The dynamic range and sensitivity was investigated by cantilever determined by X ray photoelectron spectroscopy varying the vancomycin concentration in solution [Van]. The (XPS) as described in the Supplemental Materials) to a series differential DAla bending signal scaled with increasing [Van] of [Van]. Figure 4 shows all of the stress data as a function of ~ 10, 100 and 1000 nM, giving equilibrium differential p and [Van] for both DAla and DLac. The nanomechanical signals of 8, 29 and 114 nm respectively (Figure 2c). The signal was much larger for DAla than DLac, and steeply lowest [Van] to be detected was 10 nM giving rise to a DAla increased as a function of [Van] followed by saturation, differential bending signal of 9 ± 2 nm on three cantilevers whereas it increased more gradually as a function of p. For (Figure 2d). fixed p = 1, where the cantilever is coated with only DAla, The capacity of cantilever to detect antibiotics in serum the nanomechanical response saturates for [Van] > 50 M was investigated in the clinically relevant concentration when most accessible vancomycin binding sites are occupied, range33 of 3 27 M. Figure 3 shows the differential signal for consistent with binding equilibrium. This effect is a measure DAla and DLac coated cantilevers upon injection of 7 M of the specific chemical interactions between the vancomycin vancomycin in serum (90 % fetal calf serum plus 10 % and the peptide. On the other hand, for fixed [Van], upon sodium phosphate buffer pH 7.4). The differential signal for increasing DAla from p = 0 to p = 0.1 no detectable nanomechanical signal was measured, whereas from p = 0.1 DAla in serum was 105 ± 4 nm and no significant bending was detected for DLac. to p = 1.0 there is an approximately linear increase. What this means is that the stress transduction is actually a collective DRUG TARGET NANOMECHANICAL PERCOLATION phenomenon, requiring a relatively large fraction of the surface to be covered so as to establish connectivity between We monitored the nanomechanical response of cantilevers chemically transformed regions of the surface. Assuming that with systematically varied peptide densities p (where p is the local chemistry and geometric effects responsible for the
4 Table 1 Equilibrium dissociation constant Kd and saturation stress CONCLUSIONS signals a of vancomycin-mucopeptide interactions on cantilever arrays compared with SPR and solution-phase UV spectroscopy Our experiments show that cantilever arrays have the measurements from the literature. sensitivity to detect and quantify the binding affinity of the Mucopeptide a Kd cantilever Kd SPR Kd solution (mN/m) ( M) ( M) ( M) antibiotic vancomycin to drug target mucopeptide analogues: DAla 29.7 ± 1.0 1.0 ± 0.3 1.1 ± 0.1 0.7 ± 0.1 Lysine D Alanine D Alanine and Lysine D Alanine D Lac. (ref. 13) (ref. 21) The former occurs in the peptidoglycan precursors found in DLac 14.1 ± 3.0 800 ± 310 526 ± 139 660 vancomycin sensitive bacteria and the latter in those (ref. 13) (ref. 25) precursors in VanA and VanB vancomycin resistant bacteria16 The values stated are mean ± standard error for more than 100 cantilever measurements. 27 Literature data also quoted in these units. No error analysis is given in ref. 11. . Differential measurements could successfully discriminate between the two peptide sequences, detecting the deletion of a single hydrogen bond from the drug binding pocket, which is collective build up of strain are separable, we can write a associated with drug resistance. This gave rise to an 800 fold general product form for the cantilever response, increase in Kd of DLac compared to DAla, in agreement with 11 13 α measurements made by SPR . a ⋅[Van] p-p σ = c for p > p . (2) We find that the minimum detectable vancomycin eq , c Kd + [Van] 1-pc concentration was 10 nM and comparable to SPR studies, and zero otherwise. The first term is the Langmuir which have reported the detection of 30 nM vancomycin11 13. Adsorption Isotherm, accounting for drug target binding Furthermore, we show that cantilevers can detect and quantify events and the second term is the power law form describing vancomycin in blood serum at clinically relevant the large scale mechanical consequences of stressed network concentrations, which is important for pharmacokinetic formation. The constant a corresponds to maximum surface /dynamic drug profiling, personalized medicine and forensic stress when all the accessible binding sites are occupied and applications. Kd is the surface equilibrium dissociation constant on the The molecular binding events occurring between cantilever. The build up of surface stress follows from the vancomycin in solution and DAla were found to generate a connectivity of the chemically transformed network as well repulsive compressive surface stress. The origin of the as the interactions between nodes of the network34. The biochemically induced surface stress is the subject of much 1 9 exponent of the power law α is associated with the elastic scientific debate and interest . Our experiments reveal a finite interactions between chemically reacted regions on the percolation threshold pc = 0.075 below which the macroscopic cantilever. For short range interactions, such as steric bending is effectively zero. This means that a critical number neighbour neighbour repulsive interactions there will be a of DAla and vancomycin binding events are required to yield finite percolation threshold pc beyond which there will be a observable stress and demonstrates a local short range connected network which can produce an apparent bending transduction mechanism. For p > pc, the nanomechanical of the cantilever. On the other hand, for long range signal increases where the exponent in the power law α is 1.3 interactions such as ideally elastic interactions, pc=0. and so is approximately linearly proportional to the number of We have carried out a series of least squares fits of DAla molecules on the cantilever. Figure 5 shows the Equation 2 to our data to find the key parameters as well as operation of this mechanism, which begins with the steric to ascertain the validity of the product form. The parameters forces generated by insertion of the vancomycin into the DAla pc and α, which characterize the collective behavior are SAM. The resulting complexes will induce a local strain in the coupled in a statistical sense. Therefore, to determine what silicon as well as carry an electrostatic charge, which in the their best values might be, we have chosen not to rely on the neutral pH conditions of this study is +1 for vancomycin. As multi parameter fitting routine but instead have examined the number of such regions grows they will interact to produce directly how the fit changes as a function of pc, looking at bending of the entire cantilever. both the resulting values of α and squared deviation between The proposed percolative mechanisms of antibiotic the data and the fit. Figure 4b shows the outcome, which mucopeptide triggered changes in surface stress differ reveals a percolation threshold p = 0.075, and a concomitant significantly from previous studies of the Young’s modulus of c 35,36 preference for a power α close to 1.3. We were able to macroscopic two dimensional model random elastic media . ascertain the validity of the product form, Equation 2, by Here we establish nanomechanical cantilever biosensors as the comparing the global fit values for K and a, or α and p , hosts for new universality classes of percolating systems, d c where the elasticity resides in the coupled multilayered, with the values obtained from subsets of the data at constant 37 p or constant [Van]. The analysis shown in the Supplemental multiscale system Material shows that the values do not vary outside the buffer/Van/DAla/PEG/gold/chromium/silicon. Our findings experimental error limits, and that therefore, within the errors suggest that the structure and mechanics of the underlying of our present experiments, local drug target chemical DAla/PEG SAMs play a major role in mechanotransduction. Future work will investigate the extent of mixed monolayer interaction effects decouple from the collective elastic 38,39 phenomenon ultimately responsible for the bending of the phase separation on cantilevers and the decoupling of entire cantilever. Having established this, we can find K , a, chemical and geometric factors (manuscript in preparation d MW, RAM & GA). These findings will aid rational design of α, and pc using all available data. The outcome of the global fit is superimposed onto the measured differential stress novel devices and surface chemistries for improving the sensitivity of cantilevers to chemical binding events such as signals in Figure 4a. Table 1 shows a summary of a and Kd for both DAla and DLac peptides and makes the comparison those in our current drug testing application. Interestingly we with reported values from solution phase UV show that the maximum stress signal is obtained at high DAla spectroscopy20,21 and SPR measurements11 13. packing densities, conditions that are traditionally considered
5
Figure 5 Concepts underpinning nanomechanical antibiotic transduction. (a) A schematic to show the concept of percolation on a cantilever array; (b) A schematic to illustrate nanomechanical drug target percolation on a bacterial membrane and cell wall. 12. Rao, J. Yan, L. Xu, B. Whitesides, G.M. Using surface plasmon resonance to study the to be unfavorable for other surface sensing techniques such binding of vancomycin and its dimer to self assembled monolayers presenting D Ala D Ala. 12, 13 11 as SPR (For p=1.0 we estimate 10 DAla per cantilever, J. Am. Chem. Soc. 121, 2629 2630 (1999). 2 13. Cooper, M.A. Fiorini, M.T. Abell, C. & Williams, D.H. Binding of vancomycin group with a single footprint of 44 Å , as described in the XPS antibiotics to D alanine and D lactate presenting self assembled monolayers. Bioorg. Med. Chem. 8, 2609 2616 (2000). measurements detailed in the Supplemental Material). Thus, 14. Vettiger, P. et al. Ultrahigh density, high data rate NEMS based AFM data storage system. beyond our key result that cantilevers are a very useful tool in Microelec. Eng. 46, 11 14 (1999). 15. Despont, M. Drechsler, U. Yu, R. Pogge H.B. & Vettiger, P. Wafer scale microdevice antibiotic research, our systematic experiments on model transfer/interconnect: Its application in an AFM based data storage system. J. MEMS. 13, 895 901 (2004). SAMs provide a new framework for understanding and 16. Williams, D.H. The glycopeptide story How to kill the deadly 'superbugs'. Nat. Prod. Rep. eventually engineering the response of cantilevers to 13, 469 477 (1996). 17. Williams, D.H. Maguire, A.J. Tsuzuki, W. & Westwell, M.S. An analysis of the origins of a biochemical signals. cooperative binding energy of dimerization. Science 280, 711 714 (1998). 18. Barna, J.C. & Williams, D.H. The structure and mode of action of glycopeptides antibiotics We close by speculating that nanomechanical percolation of the vancomycin group. Annu. Rev. Microbiol. 38, 339 357 (1984). may play an important role not only in sensor response but 19. Kahne D, Leimkuhler C, Wei L, & Walsh C. Glycopeptide and lipoglycopeptide antibiotics. Chem. Revs. 105, 425 448 (2005). also in the glycopeptide antibiotic mode of action in real 20. Perkins, H.R. Specificity of combination between mucopeptide precursors and vancomycin or ristomycin. Biochem. J. 111, 195 205 (1969). bacteria. Williams and co workers proposed that 21. Nieto, M. & Perkins, H. R. Modifications of acyl d alanyl d alanine terminus affecting glycopeptide antibiotic dimerisation impacts on complex formation with vancomycin. Biochem. J. 123, 773 787 (1971). 22. Mitchell, L.C. Changing patterns of infectious diseases. Nature 406, 762 767 (2000). antimicrobial efficacy and that antibiotic dimers show 23. Walsh, C. Molecular mechanisms that confer antibacterial drug resistance. Nature 406, 775 16 18 781 (2000). unusually strong binding constants . We propose further 24. Neu, H.C. The crisis in antibiotic resistance. Science 257, 1064 1073 (1992). that drug target binding events act collectively to disrupt the 25. Bugg, T.D.H. et al. Molecular basis for vancomycin resistance in enterococcus faecium BM4147 biosynthesis of a depsipeptide peptidoglycan precursor by vancomycin resistance mechanical properties of the bacterial peptidoglycan cell wall proteins VanH and VanA. Biochem. 30, 10408 10415 (1991). 26. Hopwood, D. A call to arms. Nat. Rev. Drug Discov. 6, 8 12 (2007). and plasma membrane (Figure 5b). Dimeric glycopeptides 27. Ge, M. et al. Vancomycin derivatives that inhibit peptidoglycan biosynthesis without with membrane insertive hydrophobic tails such as binding D Ala D Ala. Science 284, 507 510 (1999). 18 28. Prime, K.L. Whitesides, G.M. Self assembled organic monolayers – model systems for biphenylchloroeremomycin and oritavancin , may possess studying adsorption of proteins at surfaces. Science 252, 1164 1167 (1991). 29. Prime, K.L. & Whitesides, G.M. Adsorption of proteins onto surfaces containing end greater potency than vancomycin because of their propensity attached oligo(ethylene oxide) – a model system using self assembled monolayers. J. Am. Chem. Soc. 115, 10714 10721 (1993). to self associate and diffuse in the membrane to create 30. Lahiri, J. Isaacs, L. Tien, J. & Whitesides, G.M. A strategy for the generation of surfaces extended networks. Our results should therefore clearly presenting ligands for studies of binding based on an active ester as a common reactive intermediate: A surface plasmon resonance study. Anal. Chem. 71, 777 790 (1999). motivate future work to investigate the percolation of drug 31. Stoney, G. G. The tension of metallic films deposited by electrolysis. Proc. Roy. Soc. London, Ser. A. 82, 172 175 (1909). action in lipid membranes, including eventually those of 32. Brantley, W. A. Calculated elastic constants for stress problems associated with living bacteria, and test the more general validity of the semiconductor devices. J. Appl. Phys. 44, 534 535 (1973). 33. Rotschafer, J. C. et al. Pharmacokinetics of Vancomycin: Observations in 28 Patients and hypothesis (underlying Equation 2) of decoupling between Dosage Recommendations. Antimicrob. Agents Chemother. 22, 391 394 (1982). 34. Stauffer, D & Aharony, A. Introduction to Percolation Theory. 2nd Edition Taylor and chemical binding and geometric factors. Francis, Bristol (1991). 35. Benguigui, L. Experimental study of the elastic properties of a percolating system. Phys. Rev. Lett. 53, 2028 2030 (1984). 36. Sieradzki, K & Li, R. Fracture behavior of a solid with random porosity. Phys. Rev. Lett. References 56, 2509 2512 (1986). 1. Fritz, J. et al. Translating biomolecular recognition into nanomechanics. Science 288, 316– 37. Sushko, M. L., Harding, J. H., Shluger A. L., McKendry R. A. & Watari M. Physics of 318 (2000). Nanomechanical Biosensing on Cantilever Arrays. Adv. Mat. in press (2008). 2. McKendry, R. A. et al. Multiple label free biodetection and quantitative DNA binding 38. Folkers, J. P.,Hibinis, R. E. & Whitesides, G. M. Phase Behavior of Two Component Self assays on a nanomechanical cantilever array. Proc. Natl. Acad. Sci. USA, 99, 9783 9788 Assembled Monolayers of Alkanethiolates on Gold. J. Phys. Chem. 98, 563 571 (1994). (2002). 39. Smith, R. et al Phase Separation within a Binary Self Assembled Monolayer on Au{111} 3. Savran, C.A. Knudsen, S. M. Ellington, A. D. & Manalis, S.R. Micromechanical detection of Driven by an Amide Containing Alkanethiol. J. Phys. Chem. B, 105, 1119 1122 (2001). proteins using aptamer based receptor molecules. Anal. Chem. 76, 3194 3198 (2004). Supplemental Material accompanying this paper is provided on the following pages. 4. Calleja, M. et al. Highly sensitive polymer based cantilever sensors for DNA detection. Ultramicrosc. 105, 215 222 (2005). 5. Shu, W. et al DNA molecular motor driven micromechanical cantilever arrays. J. Am. Chem. Soc. 127, 17054 17060 (2005). Acknowledgements 6. Watari, M. et al. Investigating the molecular mechanisms of in plane mechanochemistry on We thank the Engineering and Physical Sciences Research Council (Speculative Engineering cantilever arrays. J. Am. Chem. Soc. 129, 601 609 (2007). funding programme JWN, RMcK, GA, MW, DZ,), Interdisciplinary Research Centre in 7. Backmann, N. et al. A label free immunosensor array using single chain antibody fragments. Nanotechnology (Cambridge, UCL and Bristol: MW, JWN, RMcK), the Royal Society (RMcK, Proc. Natl. Acad. Sci. USA 102, 14587 14592 (2005). JWN), Biotechnology and Biological Science Research Council (RMcK), the Wolfson 8. Zhang, J. et al. Rapid and label free nanomechanical detection of biomarker transcripts in Foundation (GA), Cancer Research UK (ADB) for funding. We thank Emily Smith (University human RNA. Nature Nanotech. 1, 214 – 220 (2006). of Nottingham EPSRC XPS Open Access scheme), Jane and Rex Galbraith (UCL Statistics) for 9. Wu, G. Datar, R. H. Hansen, K.M. Thundat, T. Cote, R.J. & Majumdar, A. Bioassay of helpful discussions. prostate specific antigen (PSA) using microcantilevers. Nature Biotech. 19, 856 – 860 (2001). Author contributions 10. Cooper, M.A. Advances in membrane receptor screening and analysis. J. Mol. Recognit. 17, JWN and MW contributed equally to this manuscript. All authors discussed the results and 286 315 (2004). commented on the manuscript. 11. Rao, J. Yan, L. Lahiri, J. Whitesides, G.M. Weis, R.M. Warren, H.S. Binding of a dimeric derivative of vancomycin to L Lys D Ala D Lactate in solution and at a surface. Chem. & Biol. 121, 353 359 (1999). Author Information Correspondance and requests for materials should be adressed to R.A.M.
6 Supplemental Material
‘Nanomechanical detection of antibiotic-mucopeptide binding in a model for superbug drug resistance’ J. W. Ndieyira, M. Watari, A. Donoso Barrera, D. Zhou, M. Vögtli, M. Batchelor, M. A. Cooper, T. Strunz, M. A. Horton, C. Abell, T. Rayment, G. Aeppli and R. A. Mckendry*.
* email: [email protected] Synthesis of thiolated peptides
The peptides DAla and DLac were synthesized by solid phase methodology using commercially available pre loaded Wang D Ala and Wang D Lac resins and standard Fmoc protecting group chemistry.S1 The cleaved products were further purified by reverse phase HPLC by varying the mobile phase from 5% to 95% of acetonitrile in water (with 0.5% trifluoroacetic acid). The peptides were characterized by NMR and HRMS (+ESI, Q TOF) as follows; i) D Ala OH (HS(CH2)11(EG)3 OCH2 Ahx L Lys D Ala D Ala OH 1 H NMR (δ ppm, 400 mHz, CD3OD): 1.25 1.80 (m, 36H, [15CH2 + 2 CH3]), 1.91 (s, 3H, CH3), 2.24 (t, 2H, CH2), 2.47 (t, 2H, J = 7.1 Hz, CH2SH), 3.14 (t, 2H, J = 7.0 Hz, NHCH2), 3.22 (t, J = 7.0 Hz, 2H, NHCH2), 3.44 (t, 2H, J = 6.7 Hz, CH2(OEG) ), 3.54 3.69 (m, 12H, 3(EG)), 3.96 (s, 2H, OCH2C=O), 4.21 (t, 1H, J = 7.1 Hz, L lys α CH), 4.31 4.42 (m, 2H, 2 [D Ala α CH]). 13 C NMR (δ ppm, 125 mHz, CDCl3): 176.24, 175.58, 174.46, 174.29, 173.21, 172.61 (6 C=O), 72.39, 71.95, 71.55, 71.52, 71.37, 71.22, 71.16 (PEG), 55.15, 50.00, 40.17, 39.83, 36.50, 35.22, 32.27, 30.72, 30.70, 30.64, 30.56, 30.27, 30.21, 29.40, 27.57, 27.21, 26.46, 24.97, 24.23, 22.57, 18.01, 17.50. + HRMS (+ESI, Q TOF), found 842.4941; required for C39H73N5O11SNa [M+Na] , 842.4925, dev. 1.86 ppm. ii) D Lac OH (HS(CH2)11(EG)3 OCH2 Ahx L Lys D Ala D Lac OH 1 H NMR (δ ppm, 500 MHz, CD3OD): 1.25 1.80 (m, 36H, [15CH2 + 2CH3]), 1.92 (s, 3H, CH3), 2.24 (t, 2H, J = 7.0 Hz, CH2), 2.47 (t, 2H, J = 7.0 Hz, CH2SH), 3.13 3.17 (m, 2H, NHCH2), 3.18 3.24 (m, 2H, NHCH2), 3.46 (t, 2H, J = 6.7 Hz, CH2(OEG)), 3.55 3.68 (m, 12H, 3(EG)), 3.96 (s, 2H, OCH2C=O), 4.30 4.40 (m, 1H, L lys α CH), 4.40 4.50 (m, 1H, D Ala α CH]). 5.00 5.10 (m, 1H, D Lac α CH). 13 C NMR (δ ppm, 125 MHz, CD3OD): 175.99, 174.12, 173.34, 173.20, 172.84, 172.60 (6 C=O), 79.47, 72.40, 71.99, 71.58, 71.55, 71.39, 71.24, 71.19 (PEG), 54.37, 53.68, 40.19, 39.82, 36.73, 36.63, 35.23, 32.84, 32.22, 30.74, 30.71, 30.64, 30.57, 30.26, 30.21, 30.00, 29.85, 29.41, 27.58, 27.52, 27.22, 26.62, 26.56, 24.98, 24.17, 22.60 (CH3), 17.51 (CH3), 17.44 (CH3). + HRMS (+ESI, Q TOF), found 843.4761; required for C39H73N5O11SNa [M+Na] , 843.4765, dev. – 0.53 ppm. iii) The synthesis of the 11 mercaptoundecyl tri(ethylene glycol) alcohol (PEG) has been described.S2 The crude compound was purified using flash column chromatography on silica (10% ethanol in ethyl acetate) to yield the final product (PEG) as a colourless oil. 1H NMR (250 MHz, CDCl3, δppm): 1.20 1.37 (m, 14H, 7CH2), 1.49 1.60 (m, 4H, 2CH2), 2.46 (q, 2H, J = 7.0 Hz, HSCH2 ), 3.05 (s, br, 1H, OH), 3.40 (t, 2H, J = 7.0 Hz, CH2PEG), 3.50 3.75 (m, 12H, 13 3(OCH2CH2). C NMR (62.5 MHz, CDCl3, δppm): 72.5, 71.4, 70.5, 70.3, 69.9, 61.5, 34.0, 33.7,
7 30.5, 29.5, 29.4, 29.0, 28.8, 28.7, 28.3, 26.0, 24.5. HRMS (Q TOF, ES+), found 359.2222; required + for C17H36O4SNa [M + Na] , 359.2232. Preparation of cantilevers Cantilever arrays were fabricated by IBM Research Laboratory, Rüschlikon, Switzerland, and purchased from Veeco Instruments Inc. (Santa Barbara, CA, U.S.A.). Each Si (100) cantilever was nominally 500 m long, 100 m wide, and 0.9 m thick. Cantilever arrays were first cleaned with freshly prepared piranha solution (at ratio 1:1 H2SO4 and H2O2) for 20 mins. Piranha solution had to be handled with care as it is hazardous and can cause explosions or severe skin burns. Arrays were then thoroughly rinsed in deionized water before being immersed in the second freshly prepared piranha solution for another 20 mins, and again rinsed thoroughly with deionized water. Finally, the arrays were rinsed with pure ethanol and dried on a hotplate at 75 °C. They were then inspected using the optical microscope to confirm their cleanliness before transferring to the evaporation chamber (BOC Edwards Auto 500, U.K.) for overnight pumping. One side of each cantilever array was coated by thermal evaporation with a 2 nm Cr adhesion layer followed by 20 nm of gold from a base pressure ~ 2×10 7 mbar, and using evaporation rates of 0.02 nm/s for Cr and Au, as measured directly above the source by a quartz crystal monitor. Once the required thickness was attained, the samples were left in the chamber for 1 2 h to cool under vacuum before opening. The freshly evaporated cantilevers were sealed in a vacuum storage vessel (Agar Scientific, U.K.) filled with argon and functionalized within a few hours after the evaporation. The cantilevers were incubated in an array of eight glass microcapillaries filled in a random order with 2 mM ethanolic solutions of DAla, DLac or PEG for 20 mins, and then rinsed in pure ethanol and deionized water. The cantilevers were stored under distilled water immediately after functionalisation until use. Cantilevers were used immediately but where it was not possible to use all of them within a day, then they were kept in the Petri dishes with deionised water as we have previously found that functionalised arrays could be stored under deionised water for 2 weeks without degrading as opposed to being stored in air.
Preparation of solutions Buffer solutions were prepared by mixing 0.1 M mono and di basic sodium phosphate salts (Sigma Aldrich, U.K.) dissolved in ultrapure water (18.2 M cm resistivity, Millipore Co., Billerica, MA, U.S.A.) to yield a pH value of 7.4. Fresh buffer was used to prepare the solutions of vancomycin (Sigma Aldrich, U.K). The solutions were filtered using 0.2 m filters (Millipore) and ultrasonicated for 30 mins at room temperature before being purged with argon. To investigate the detection sensitivity of cantilevers to antibiotics in serum, vancomycin solutions were prepared using 90% fetal calf serum (Gibco, USA) with 10 % sodium phosphate buffer. Instrumentation The absolute bending of all eight cantilevers was monitored using the serial time multiplexed optical beam method with a single position sensitive detector (Scentris, Veeco Instruments Inc.). The functionalized cantilever array was mounted in a sealed liquid chamber with a volume of approximately 80 l. The liquid cell and each aliquot of aqueous solutions were placed into the temperature controlled cabinet overnight to allow for temperature equilibration before undertaking
8 each experiment. Alignment of each laser spot onto the free end of each cantilever on the array was confirmed using 1 °C temperature heating test. The purpose of the heating test was to ensure that the effective length of all eight cantilevers was similar. This was witnessed by downward bending, due to the bimetallic effect. The relative standard deviation of the absolute bending signals of all eight cantilevers was ≤ 5%. Herein a negative deflection signal denotes a compressive surface stress and a positive signal is tensile. The efficient exchange of liquids in the liquid cell was achieved with a home built gravity flow microfluidics system. The desired flow rate was first determined before each experiment to ensure a constant flow rate for all the arrays. The data acquisition was automated using LabView (National Instruments Co., Austin, TX, U.S.A.) software via a 6 way valve (Serial MVP, Hamilton, Reno, NV, U.S.A.). The measurement protocol involved the following: i) sodium phosphate buffer solution (pH 7.4, 0.1 M) for 30 mins to establish a baseline; ii) injection of vancomycin in sodium phosphate buffer for 30 mins; iii) to dissociate the complex we then injected sodium phosphate buffer (pH 7.4 , 0.1 M) wash for 30 mins; iv) a further washing step using 10 mM HCl for another 30 mins to remove the vancomycin and regenerate the peptide surface; v) finally another sodium phosphate buffer step to restore the baseline signal. This process was repeated for all the vancomycin concentrations. All signals were acquired under a constant liquid flow rate of 180 ± 30 l/min.
Validity of the product form of Equation 2 Least square fits were applied to subsets of the data, in the form of constant p or constant [Van] cuts through the data, and determining whether Kd and a or pc and α vary outside the experimental error limits with [Van] and p, respectively. The results are summarized in Table S1. The fitted parameters a = (29.7 ± 1.0) mN/m, Kd = (1.0 ± 0.3) M, pc = (0.08 ± 0.09), and α = (1.3 ± 0.3) given in the first row were obtained by including all available data; the fitted parameters a = (29.5 ± 2.0) mN/m and Kd = (0.8 ± 0.5) M at fixed pc = 0.08 and α = 1.3 given in the second row were obtained by including p = 1.0 data only; the fitted parameters pc = (0.00 ± 0.40) and α = (1.9 ± 1.3) at fixed a = 29.7 mN/m and Kd = 1.0 M given in the third row were obtained by including [Van] = 10 M data only; the fitted parameters pc = (0.09 ± 0.17) and α = (1.2 ± 0.5) at fixed a = 29.7 mN/m and Kd = 1.0 M given in the fourth row were obtained by including [Van] = 100 M data only; the fitted parameters pc = (0.17 ± 0.07) and α = (0.8 ± 0.3) at fixed a = 29.7 mN/m and Kd = 1.0 M given in the fifth row were obtained by including [Van] = 250 M data only. These results show that the fitted parameters a, Kd, and α obtained from the global fit and the fitted parameters obtained from the fit of data subsets were consistent, and that therefore, to within experimental error, local chemical interaction effects decouple from the collective elastic phenomenon ultimately responsible for the bending of the entire cantilever. Table S1: Least square fits of data subsets to validate the product form of Equation 2. p [[[Van]]] ] (((( M) a (mN/m) Kd ( M) pc ααα all(a) all(b) 29.7 ± 1.0 1.0 ± 0.3 0.08 ± 0.09 1.3 ± 0.3 1.0 All 29.5± 2.0 0.8 ± 0.5 0.08 1.3 all 10 29.7 1.0 0.00 ± 0.40 1.9 ± 1.3 all 100 29.7 1.0 0.09 ± 0.17 1.2 ± 0.5 all 250 29.7 1.0 0.17 ± 0.07 0.8 ± 0.3
(a) Including all values of p investigated herein, i.e. 0.05, 0.1, 0.3, 0.5, 0.7, 0.9, and 1.0. (b) Including all values of [Van] investigated herein, i.e. 0.1, 1, 10, 50, 100, 250, 500,750, 1000 M
9 X-ray photoelectron spectroscopy
The relation between the solution molar fraction and surface coverage of DAla was measured on the resultant SAMs by X ray photoelectron spectroscopy. Given the miniaturised geometry of cantilevers, characterisation of the SAMs was performed on silicon wafers instead, which had been functionalized in parallel with the silicon cantilevers. The data were captured using Kratos VISION II software (version 2.2.6) on a Kratos Axis Ultra spectrometer equipped with aluminum Al Kα source and 1486.6 eV line energy. The X ray spot size was 1 mm² and the analysis area was defined by the SLOT aperture of 300 × 700 m² with a hybrid (magnetic/electrostatic) optics and a multi channel plate and delay line detector (DLD) with a collection angle of 30o and a take off angle of 90o. The pressure in the chamber was 3x10 9 Torr. Preliminary wide scans were taken over the full range of 1400 to 0 eV, with step size of 1 eV and pass energy of 80 eV to improve signal to noise ratio on the peaks of interest and then reduced down to 700 to 0 eV for the main analysis of the high resolution scans on N (1s), S (2p), Au (4f), and C (1s) spectra acquired with a 20 eV pass energy using scan step size of 0.1 eV. The averaging was undertaken over three spectra in three separate areas on each sample. Figure S1 shows typical raw high resolution photoemission spectra of N (1s), S (2p), and Au (4f) acquired on a SAM prepared from the ethanolic solution at a DAla to PEG ratio of 0.50:0.50 superimposed to fitted curves. The peak fitting of the raw spectra was carried out using commercial software (CasaXPS, 2.3.2) by using the same line shape and full width half maximum on each sample, with a linear background subtraction. Errors in the fitted area under the photoemission peaks were calculated as the standard error of the mean of the difference between the raw and fitted data.
Figure S1: Typical raw high resolution photoemission spectra of (a) N (1s), (b) S (2p), and (c) Au (4f) acquired on a SAM prepared from the ethanolic solution at a DAla to PEG ratio of 0.50:0.50, superimposed to fitted curves in grey.
Langmuir adsorption models The competitive adsorption of DAla and PEG thiols onto gold can be described by the chemical reaction [DAla]solution + [PEG]surface ↔ [DAla]surface + [PEG]solution , Equation S1 where [DAla]solution and [PEG]solution is the solution concentration and [DAla]surface and [PEG]surface is the surface concentration of DAla and PEG, respectively. The equilibrium of this reaction is characterised by the equilibrium constant [DAla]surface [PEG]solution K eq = . Equation S2 [][]DAla solution PEG surface
10 In the cantilever experiments presented herein, we varied the ratio of DAla and PEG thiol solutions at a fixed total solution concentration of thiols [thiol]solution = [DAla]solution + [PEG]solution of 2 mM (dissolved in ethanol). Therefore the solution molar fraction of DAla and PEG totals to solution solution χ DAla + χ PEG = 1.0. Similarly, the total surface concentration of the thiols is [thiol]surface = [DAla]surface + [PEG]surface and the surface coverage of the thiols is surface surface surface surface χ thiol = χ DAla + χ PEG . The total surface coverage of the thiols is defined as χ thiol = 0.1 if the surface surface coverage of DAla and PEG are described relative to one another, whereas χ thiol depends surface surface on χ DAla and χ PEG if the surface coverage of DAla and PEG are related to the total adsorption sites on the gold surface, which is due to differences in the molecular volume (hence steric effects) between DAla and PEG thiol molecules. To describe the surface coverage of DAla as a function of its solution molar fraction, the binding constant of DAla KDAla is formulated by means of Equation S2 and the relations solution solution surface surface χ DAla + χ PEG = 1.0 and χ thiol = 0.1 , which is then solved for χDAla to give solution surface solution K DAla χ DAla χ DAla ()χ DAla , K DAla = solution . Equation S3 1+ χ DAla ()K DAla −1 surface surface surface surface surface The total surface coverage of thiols χ thiol (χ DAla , χ PEG ) = χ DAla + χ PEG can then be derived by combining Equation S3 with Equation S2 and solving the expression for the binding constant of surface thiols K thiol for χ PEG , which reads solution surface solution K DAla (1− χ DAla ) χ PEG ()χ DAla , K DAla , K thiol = solution Equation S4 K thiol ()1+ χ DAla ()K DAla −1 and so the total surface coverage of thiols corresponds to the sum of Equations S3 and S4.
Analysis of DAla and thiol surface coverage
surface The surface coverage of DAla peptides χDAla can be directly correlated with the intensities of N (1s) photoelectrons, since nitrogen is present in the molecular sequence specific to the DAla surface peptide only. The surface coverage of thiols χ thiol was measured as the ratio of the intensities of S (2p) and Au (4f) photoelectrons, which have the same escape depth and are thus attenuated to the same extent by the organic overlayer. Figure S2 shows the measured N (1s) intensities and the ratios of the intensities of S (2p) and Au (4f), both normalised to the 100% DAla SAM, as a function of DAla solution molar fraction and superimposed to the fit of Langmuir model Equation S2 and the sum of Equation S2 and S3, respectively. As can be seen in the figure, the relation between the measured surface coverage and the DAla solution molar fraction was well described by the Langmuir models. The fitted binding constants, K DAla = .0 25 for the adsorption of DAla peptides and K thiol = .0 15 for the adsorption of thiols suggest that the DAla peptides adsorbed slower and had a lower packing density compared to PEG molecules. For example, the thiol coverage of the 100% DAla SAM exhibited a 39% lower value compared to the 100% PEG SAM. This result is rather intuitive because of the larger molecular volume occupied by the DAla peptides as well as the increased molecular length, both giving rise to greater steric effects and diffusion length compared to more compact PEG molecules.
11 Figure S2: Normalised N (1s) intensities and the ratios of the intensities of S (2p) and Au (4f), as a function of DAla solution molar fraction and superimposed to the fit of Langmuir model. The analysis of the surface coverage allows us to evaluate the cantilever measurements performed at different DAla peptide densities and vancomycin solution concentrations. Figure S3 shows the differential stress as a function of (a) the DAla surface coverage superimposed to the percolation model (i.e. geometrical term in Equation 2) and (b) the solution vancomycin concentration superimposed to the Langmuir model (i.e. chemical term in Equation 2). The three dimensional representation of these data is presented in Figure 4 of the manuscript.
Figure S3: Measured differential surface stress response for DAla and DLac coated cantilevers as a function of (a) DAla surface coverage and (b) vancomycin concentration in solution [Van] superimposed with the results of the global fit according to Equation 2 (solid lines). Furthermore, our X ray photoelectron spectroscopy data can give an estimate of the surface area occupied by one DAla peptide molecule, by referring to the literature value of 27 Å2 per thiolate reported for PEG SAMs.S3 This translates into a molecular area of 44.2 Å2 in the 100% DAla SAM or a nearest neighbour distance of 7.1 Å by assuming an ideal hexagonal packing.
References (S1) Cho Y. R., Entress R. M. & Williams, D. H. Synthesis of cell wall analogues of vancomycin resistant enterococci using solid phase peptide synthesis. Tetrahedron Lett., 38, 5229 5232 (1997). (S2) Zhou D., Ying L., Hong X., Hall E. A, Abell C. & Klenerman, D. A. Compact functional quantum dot DNA conjugate: preparation, hybridization, and specific label free DNA detection. Langmuir, 24, 1659 1664 (2008). (S3) Schwendel D., Hayashi T., Dahint R., Pertsin A., Grunze M., Steitz R. & Schreiber F. Interaction of water with self assembled monolayers: neutron reflectivity measurements of the water density in the interface region. Langmuir, 19, 2284 2293 (2003).
12